Your research, one step at a time.

1. Open Meta-analysis opportunities. Search by topic or journal, then open a study to read more.
2. Follow View on PubMed or the journal link to read the original report. Compare its population, intervention, outcomes, and study design with your question.
3. Read the saved assessment, its Last checked date, and the sources checked. Treat a candidate label as a starting point for your own review.

Promising means source-supported compatibility was found in at least two provisionally distinct primary-study groups. Needs verification means eligibility, overlap or numerical data remain uncertain. Not currently poolable applies only to the screened evidence. Open the study-by-study checks, verify the quoted evidence against the full reports, and confirm outcome definitions and follow-up. Previous reviews are background only and never count as included primary studies. Historical assessments need a fresh check under the stronger criteria.

Next: save the study if you want to return to it or discuss a review idea.

1. Select Save to favourites to keep a private shortlist in My favourites.
2. Add names and email addresses in Contacts. Saving a contact does not share any work.
3. Use the study’s Share icon, choose a review project, write your idea, and select just one person or several. Choose whether you want feedback or collaboration, then send the invitations.

Use a new project to share with only the selected people; an existing project already has its own team. Recipients accept with the matching verified email. You can invite more people later from Team.

Know the difference: Copy puts text or a link on your clipboard. Share opens the project and invitation choices.

1. Create or open a project in Projects. Open Define your question.
2. Describe the population, intervention, comparison, outcomes, timing, and study types. Add inclusion and exclusion rules and review the final question summary.
3. Save the question. Choose Manual review · recommended under your review approach and invite independent reviewers from Team. Two human reviewers are the default.

Ready to continue when: your saved question and eligibility rules are clear enough for every reviewer to use.

1. Under Upload search results, select the database and upload its reference export. Use PubMed NBIB, MEDLINE text or XML; Embase or Cochrane RIS; or a reference CSV.
2. Open Check duplicates and detect possible matches. Compare each pair manually. An authorized assigned reviewer or project leader can confirm the manual choice of either copy. AI and source warnings advise that choice. Use Ask team leader to review when you need help. Keep separate reports of the same study and link them with Keep both · link same study. Undo my last change and removal history provide recovery. Rerun matching preserves decisions; Restart duplicate check starts over and asks you to confirm clearing duplicate decisions.
3. Open Screen titles and abstracts, read each reference, then choose Include, Exclude, or Unclear with your reason. Save decision & next reference advances your queue.
4. At Screen full texts, attach and read each eligible full report. Save your own decision, then resolve team disagreements before proceeding.

Ready to continue when: the team has agreed which full-text studies belong in the review.

1. Open an included study in Extract study data. Enter values manually for each outcome, comparison, and time point. Record the source quote and page or table.
2. Check group direction, units, sample sizes, and missing values. Save a draft while working; submit your reviewer form when you have checked it.
3. Open Assess risk of bias. Choose the appropriate tool and outcome, record your evidence and judgments, and submit the assessment.

Ready to continue when: the required reviewers have submitted extraction and risk-of-bias forms backed by source evidence.

1. A project owner or manager compares submitted forms, checks discrepancies against the papers, and approves the reconciled extraction and risk-of-bias data.
2. Open Review your analysis plan. Choose an outcome, follow-up time, and comparison, then select Preview final dataset.
3. Inspect every row. Download the approved CSV if you want to review it separately. When ready, a manager selects Create R analysis draft.

Next: review the draft’s analysis choices in step 7. Creating a draft does not start R.

07

REVIEW AND RUN

Confirm your choices, then start R.

1. Check your data. Review study names, headings, and numbers. Use “Are the column headings on the wrong row?” if needed.
2. Choose whether to use AI. On the first review screen, keep manual choices or select automatic smart-hosted classification. Automatic classification runs after your selected outcomes pass the basic checks; you still approve or edit every proposal.
3. Choose outcomes. Select the outcomes and data type. Check or correct the study and group columns.
4. Read the basic R analysis. Continuing from outcome selection fits a short baseline using your choices. Review every included study and value, the pooled result, and the available or unavailable subgroup and meta-regression candidates. Change settings and rerun the basic analysis when needed; it does not use a final-analysis credit.
5. Keep AI optional. Use Continue on the optional AI screen to review your own choices when no suggestion is pending. If you request a suggestion, check the proposed methods, moderator columns, and treatment labels against your protocol. Selected suggestions remain editable before approval.
6. Read Planned calculations and why. Review the effect measure, pooling, model assumptions, uncertainty calculation and selected moderators. Each AI setting also explains why it was proposed and what you must confirm.
7. Review Your analysis summary. Choose “Yes, edit my choices” if anything needs changing. Confirm “No more edits” only after checking the summary.
8. Run this analysis. R calculates the statistics. Follow the processing status, or return later through My analyses.

What uses your allowance: starting a run or rerun uses an analysis. Saving a draft does not. AI usage is logged separately.

08

REVIEW THE RESULTS

Check, interpret, and keep your work.

1. Read the data checks, warnings, and analysis notes first. A completed run means the calculation finished; you still need to review the methods and conclusions.
2. Open All result plots to see every generated displayable image. Compare the report and plots with the settings you approved. Check effect direction, uncertainty, heterogeneity, and sensitivity results.
3. Download the report, statistical summary, results package, and saved settings before the displayed expiry date. Review-project teammates can download completed project results from the project’s analysis page.

Need to compare choices? “Try again with different settings” creates a new draft and keeps the original analysis. Running the new draft uses another analysis.